Abstract

Diarrheal diseases are a global public health burden, killing 1.8 million people annually. Diarrhea disproportionately affects children and those in poverty. Most diarrheal cases can be prevented through safe drinking water, basic hygiene and/or sanitation measures, with drinking water interventions having the most impact on reducing diarrheal disease. There is no generally agreed-upon field method for determining biosand water filter effectiveness that is usable in low-resource communities. A pilot study was conducted of potential field use indicators, including the Colilert coliform presence/absence (P/A) test, hydrogen sulfide, alkalinity, hardness, pH, and fluorescently labeled latex microspheres. The study included both laboratory and field testing. The Colilert P/A test had the highest correlation to the United States Environmental Protection Agency standard method (IDEXX Quanti-trays), but more data are needed before making a recommendation. This study adds to understanding about evaluation of biosand water filters and provides preliminary data to address the need for a field use indicator for biosand water filters.

INTRODUCTION

Description of the problem

Diarrhea and other water-related diseases are a global public health burden, killing 1.8 million people annually (Prüss-Ustün et al. 2014). Diarrhea disproportionately affects children and those in poverty. Globally, diarrhea is the second leading cause of death of children age five and under, with approximately 760,000 children dying annually (WHO 2013). Ninety percent of diarreal deaths occur in developing nations (Gill et al. 2012); it is the primary cause of childhood morbidity and mortality in sub-Saharan Africa (O'Reilly et al. 2012).

Biosand water filters

One point-of-use method to improve drinking water and reduce disease is the biosand water filter (BSF), created by Dr. David Manz in the 1990s at the University of Calgary and promoted by his organization, Centre for Affordable Water and Sanitation Technology (CAWST) (CAWST 2009). An adaptation of the traditional slow sand filter, it was made smaller and modified for intermittent use, for utilization in households globally (CAWST 2009).

BSFs (structure displayed in Figure 1) are composed of a concrete or plastic container and layers of sand and rock prepared using CAWST methods (CAWST 2009). Hydraid and Samaritan's Purse organizations widely distribute versions with a plastic filter body (Hydraid 2014; Samaritan's Purse 2015). It improves drinking water using four methods: mechanical trapping, predation, adsorption, and natural death (CAWST 2009).

Figure 1

Biosand filter structure (CAWST 2009).

Figure 1

Biosand filter structure (CAWST 2009).

Studies have shown 27–74% reduction in incidence of diarrheal disease with BSF use (O'Connell et al. 2017a). BSFs have been shown to remove significant amounts of Escherichia coli in both laboratory and real-world settings ranging from 48% to 100%, with fecal coliform removal of 33.7–96.1% (O'Connell et al. 2017a). Further, BSFs have also been shown to remove echovirus type 12 (93%) and bacteriophages (82%) (Elliot et al. 2008). Studies indicate a high long-term user acceptance in communities ranging from 77% to 94% (Earwaker 2006; Fiore et al. 2010; Aiken et al. 2011; Stauber et al. 2011; Mangoua-Allali et al. 2012). BSF life span is shown to vary widely, ranging from under a year to over 12 years (Sisson et al. 2013). Therefore, it is necessary to test the filters over time to determine if they are working effectively. However, in rural developing communities where BSFs are often used, it is often not feasible to use standard laboratory testing due to limitations in electricity, transportation times, and funding. Currently, studies do not point to an agreed-upon field method for testing (O'Connell et al. 2017a).

Study community

Cyegera is a small community in Southern Province, Rwanda. Diarrhea causes 17% of total deaths in Rwanda (Prüss-Ustün et al. 2014). Demographic and Health Surveys 2010 data show that 13% of children under 5 years old had diarrhea in the 2 weeks preceding the survey (DHS Program 2015). The health workers at the clinic in Cyegera reported that diarrhea and intestinal parasites are of major concern in the community, especially in children (personal communication, July 2014). Data from the local Compassion International branch show that 2.13% of the children, aged 6–14 years, enrolled in the program had diarrhea from July 2014 to June 2015 (personal communication, June 2015). Diarrhea was second to malaria in causing illness in children enrolled in the program and it continues to be a problem for older children, but the greatest risk is in children under five. Health data specific to the community of Cyegera are not available other than through personal communication with local authorities.

Study aim

The specific aim of this study was to carry out a pilot laboratory trial and a field trial in Cyegera of selected potential field use indicators of biosand water filters and to obtain data for further study.

METHODS

Laboratory testing of field use indicators

To address the gap in the literature of evidence-based methods usable in a remote field setting to determine whether a BSF is functioning, a pilot study was conducted of six existing potential field use indicators (FUIs): Colilert's presence/absence (Colilert P/A), Hach's hydrogen sulfide, alkalinity and hardness kits, a Mettler Toledo EL-2 battery-powered pH meter, and Sigma Aldrich fluorescently labeled latex microspheres. Laboratory procedures were approved by the East Tennessee State University (ETSU) Biosafety Committee.

Description of FUIs and standard

IDEXX Colilert Quanti-trays were used as the standard for comparison of the FUIs. The system provides most probable number (MPN) counts for coliforms (IDEXX 2015a). The number of trays that change color to yellow are counted and the MPN table used to determine results (IDEXX 2015a). IDEXX recommends incubating (IDEXX 2015b), but for this study, incubation was not done so that results would be replicable in the field. Thus, results (color change) were read at 12-hour increments for 48 hours, rather than only at the recommended 24-hour interval.

The Hach alkalinity test kit determines the total alkalinity of the water sample (Hach 2015). The Hach hardness procedure was used to determine hardness of the sample.

A Mettler Toledo EL-2 battery-powered digital pH meter was used to test pH of each sample.

Two types of Sigma Aldrich latex microspheres were used: 2.0 μm yellow-green and 1.0 μm red. At first use, 1 mL of each size was mixed in the influent to be put in the four filters in treatment group A. The microspheres were used only in treatment group A because microspheres or microbeads are sometimes used in filter media as a filtration aid and could, therefore, potentially affect the filtration process (Balsimo & Mary 1994).

Study design

Laboratory evaluation of the FUIs occurred using nine BSFs constructed during August and September 2014 at ETSU. The filters were segregated into three groups: four filters receiving microspheres in the first influent (group A), four filters not receiving microspheres (group B), and one control (C). Group A and group B were established to delineate whether the microspheres affected filter operation or parameter measurement because they were being poured directly into the BSFs. The control BSF was used to detect any environmental fluctuations influencing the testing. Unamended tap water was used for the control. Escherichia coli was added to the influent so that there was a known contaminant for testing.

The E. coli was isolated from an environmental sample and tests were done to ensure that it was not of the O157:H7 serotype. The sample was initially cultured on Eosin Methylene Blue (EMB) agar plates, and then transferred to slants of Tryptic Soy agar. To prepare E. coli for addition to the influent, 300 mL of Tryptic soy broth was inoculated with a single loop of E. coli and incubated at 35 °C for 18 hours. This broth was divided into 100 mL increments and centrifuged for 10 minutes at 800 RCF (relative centrifugal force), producing an E. coli ‘pellet’ at the base of each centrifuge container, above which the broth was removed. Two mL of phosphate-buffered saline (PBS) was then added, the mixture vortexed, and this solution was transposed for influent seeding.

Subsequently, 2,000 mL samples were collected from each filter, and 200 mLs of each sample was taken for the Colilert Quanti-tray test and for the microspheres. The Quanti-tray was conducted according to the manufacturer's procedures.

Field trial of field use indicators

Researchers traveled to Cyegera, Rwanda, during 4–30 June 2015 to test the FUIs under real-world conditions. It was a cross-sectional study intended for future scale-up and repetition to become a longitudinal study. ETSU and University of Rwanda Internal Review Boards and the Rwanda National Ethics Committee did not require the FUI data collection procedures to be reviewed because it was not considered human subjects' research.

BSF installation

For the field trial, eight new BSFs constructed by the Rwandese Health and Environment Project Initiative (RHEPI) were installed in homes throughout the village near water sources, and two existing BSFs in the children's home were re-installed. The project was financed through an ETSU Research Development Committee Major Grant. The process of BSF installation was described using IBM-WASH psychosocial and technological factors at all levels (Dreibelbis et al. 2013). The BSFs were intended for community access due to budget constraints, but were installed in individual homes so that there was a sense of individual responsibility for maintenance and reporting of any problems.

Sample collection and FUI procedures

Water samples were collected from each filter at installation and on 27 June 2015. It was intended that the second set of samples would be taken 2 weeks after installation to allow the biolayer to form; however, samples were tested slightly earlier. This was due to schedule delays in installing new BSFs and time troubleshooting the older BSFs. Further, the FUIs needed 48 hours to process. Samples were collected and tested from the four community water sources on 22 June 2015. FUI procedures from laboratory testing were followed in the field with one exception – the Colilert P/A test was read every 6 hours instead of 12 due to possible ambient air temperature differences.

In July 2017, 2 years after installation, the BSFs were visited again. One had been destroyed in an accident in which it was tipped over. Another had been discontinued due to the owner moving away. New residents of the house where the filter was not in use were educated and said they were eager to utilize it. Samples from these filters were taken on 24 and 26 July 2017, and tested with Colilert P/A.

In May 2018, follow-up revealed that filters installed in 2015 were all in use. One had a rusty diffuser plate, which was replaced. Water from existing filters and sources was sampled on 1 June 2018. Samples were tested with Colilert P/A.

Data analysis

Laboratory data and field data were entered into Microsoft Excel spreadsheets. The method detection limit as provided by the manufacturers was used when results were below the detection limit (IDEXX Colilert Quanti-tray – one organism per 100 mL, Colilert P/A test – two heterotrophs per 100 mL, Hach hydrogen sulfide kit – 0–5 mg/L, Hach hardness – 1–20 mg/L, and Hach alkalinity – 5–100 mg/L). All data were imported into SPSS version 23.0.

Transformations were performed on the results of each test to make the data as normally distributed as possible to meet requirements for use of statistical testing. Data for pH and Colilert P/A were used without transformation. Logarithmic transformation produced the most normal data for Quanti-tray MPN and Hach hardness tests. Square root transformation produced the most normal data for Hach alkalinity and hydrogen sulfide tests. Pearson's R statistic was performed for each FUI compared to Quanti-trays to determine correlation. Field trial data were reported for the one FUI that had correlation to log Quanti-tray MPN.

RESULTS

Laboratory evaluation

The microbeads that were put into the first influent were never seen in effluent samples collected. Because no difference was indicated in the filter group with microbeads, results from filter groups A and B were combined for analysis. Negative values for the Colilert P/A at 12 or 24 hours prevented correlation analysis for these time periods. Pearson's R correlation is provided in Table 1 for each test compared to Quanti-tray results.

Table 1

Pearson's R statistic results – categorical log MPN for each FUI

Field use indicator Pearson's R compared to categorical log Quanti-tray MPN 
Colilert P/A at 36 hours 0.642 
Colilert P/A at 48 hours 0.503 
pH meter −0.037 
Hach hardness (log) −0.014 
Hach alkalinity (square root) −0.075 
Hach hydrogen sulfide (square root) 0.151 
Field use indicator Pearson's R compared to categorical log Quanti-tray MPN 
Colilert P/A at 36 hours 0.642 
Colilert P/A at 48 hours 0.503 
pH meter −0.037 
Hach hardness (log) −0.014 
Hach alkalinity (square root) −0.075 
Hach hydrogen sulfide (square root) 0.151 

Quanti-tray MPN data were categorized per the WHO recommended acceptable drinking water concentration for E. coli in drinking water, which is <10 MPN per 100 mL (WHO 2011). This corresponds to <1 log MPN. Categorical MPN results were compared to the categorical Colilert P/A results using chi-square. The null hypothesis for the chi-square tests was that there was no significant difference between results of the risk-categorized MPN and Colilert P/A results. Chi-square results for the categorical log Quanti-tray MPN and Colilert P/A are displayed in Tables 2 and 3.

Table 2

Chi-square Colilert P/A 36 hour

  Colilert P/A 36 hour
 
 
  Total 
Log Quanti-tray MPN categorical ≤1 log MPN/100 mL 41 11 52 
>1 log MPN/100 mL 15 18 
 Total 44 26 70 
  Colilert P/A 36 hour
 
 
  Total 
Log Quanti-tray MPN categorical ≤1 log MPN/100 mL 41 11 52 
>1 log MPN/100 mL 15 18 
 Total 44 26 70 
Table 3

Chi-square Colilert P/A 48 hour

  Colilert P/A 48 hour
 
 
  Total 
Log Quanti-tray MPN categorical ≤1 log MPN/100 mL 29 23 52 
>1 log MPN/100 mL 17 18 
 Total 30 40 70 
  Colilert P/A 48 hour
 
 
  Total 
Log Quanti-tray MPN categorical ≤1 log MPN/100 mL 29 23 52 
>1 log MPN/100 mL 17 18 
 Total 30 40 70 

The χ2 for Table 2 was 22.143 (p ≤ 0.001). Because one cell had a value less than five, it was necessary to use Fisher's exact test, which was also significant (p ≤ 0.001). Type II error was calculated at 0.068 and type I error at 0.423. The χ2 for Table 3 was 13.767 (p ≤ 0.001). Again, Fisher's exact test was significant (p ≤ 0.001). Type II error was calculated at 0.033 and type I error at 0.575.

Field trial results

In the field trial, none of the samples collected changed color for the Hach alkalinity test. Further, results for Colilert P/A readings under 36 hours could not be used because correlation to Quanti-trays could not be established from the laboratory data and results. Because correlation to Quanti-trays was low for other tests, field trial results were analyzed for Colilert P/A readings. Colilert P/A test was the recommended field use indicator based on the laboratory testing.

In 2017, samples from the seven BSFs in use were tested using Colilert P/A and results were negative at 36- and 48-hour readings. A sample from the BSF not in use was also taken and tested. Results indicated the presence of fecal coliforms at 36 hours, probably because the biolayer was not formed yet. A water source was tested for comparison and positive for fecal coliforms at only 24 hours.

In 2018, samples from the existing BSFs and source waters were tested using Colilert P/A. All water before filtration resulted in positive readings at 24 hours. Two samples after filtration also tested positive at 24 hours. One of these was from the filter that had a rusted diffuser plate which was replaced. The other positive sample was from a filter, which was resampled, yielding a negative result through 48 hours of incubation. The other six samples from BSFs were negative at 24 hours but positive at 36 hours. An important difference between sampling events in 2017 and in 2018 is season. In May 2018, the rainy season was still underway, likely impacting contamination levels of source waters. Table 4 displays test results of samples taken from drinking water sources, from BSFs at installation, and from BSFs on 27 June 2015, July 2017, and May 2018.

Table 4

Field trial Colilert P/A results

  Incubation time and number of samples
 
  24 hours
 
36 hours
 
48 hours
 
Date Source Negative Positive Negative Positive Negative Positive 
June 15, 2015 Water sources 
June 15, 2015 Filters at installation 
June 27, 2015 Filters 
July 26, 2017 Water source 
July 26, 2017 Filters 
May 31/June 12, 2018 Water sources 
May 31, 2018 Filters 
  Incubation time and number of samples
 
  24 hours
 
36 hours
 
48 hours
 
Date Source Negative Positive Negative Positive Negative Positive 
June 15, 2015 Water sources 
June 15, 2015 Filters at installation 
June 27, 2015 Filters 
July 26, 2017 Water source 
July 26, 2017 Filters 
May 31/June 12, 2018 Water sources 
May 31, 2018 Filters 

DISCUSSION

Laboratory testing of FUIs

Interpretation of laboratory evaluation of the FUIs should include consideration of the wide range of corresponding Quanti-tray MPN to the Colilert P/A results. The positive results for 36-hour readings of the P/A test corresponded to log MPN values of <0 to 3.38. The negative results for 36-hour readings correlated to log MPN values ranging from <0 to 2.21. Three of the negative results for Colilert P/A, out of 70 testing dates, corresponded to log MPN values that are above the WHO value, and six of the positive results for Colilert P/A corresponded to log MPN values within the WHO recommendation.

Further testing and more data would potentially enhance the predictive value and reduce the type I and II errors. Reduced correlation between Quanti-tray MPN and 48-hour readings of Colilert P/A was likely because positive results for Colilert P/A at 48 hours corresponded to a broader range of Quanti-tray MPN. While only one negative P/A result corresponded to an MPN value above one, positive P/A results corresponded about half of the time to MPNs within acceptable risk. While risk of a type II error, which would result in recommending use of unsafe water, is reduced with the 48-hour reading of Colilert P/A, the risk of a type I error is increased. A type II error is more dangerous in terms of preventing diarrhea, but a type I error, which would result in not recommending use of safe drinking water, could be detrimental in communities with limited water quantity.

Field trial of FUIs

Several factors are important to note when interpreting the results of the field trial. The 27 June 2015 results were obtained 12 days after installation of eight of the BSFs, 9 days after the re-installation of another, and only 7 days after re-installation of the final BSF. Therefore, the final testing date did not allow full formation of the biolayer. Further, because the Colilert P/A test was conducted without incubation or other temperature controls, the laboratory results and field trial results likely differ because of ambient temperature. However, the results from July 2017 indicate that seven BSFs were improving water quality compared to unfiltered water after 2 years of use. Data from May 2018 indicate that filters were less effective in removing contaminants than in 2017, but were reducing contaminants from source water. These results are likely influenced by collection during the rainy season, which is shown in studies to impact water quality overall (Prüss-Ustün et al. 2014; O'Connell et al. 2017b).

LIMITATIONS

Important limitations included failure to collect and test multiple samples per filter on testing days for purposes of reliability analysis, and small sample sizes for both laboratory testing and the field trial. Additionally, although source waters were tested for comparison to filtered samples in the field trial, results would be strengthened by testing samples directly before and after filtering. Further, summaries of field testing were based on the significant assumption that Colilert P/A tests could be used to estimate risk in the same way that Quanti-trays could.

RECOMMENDATIONS

More study is needed both in laboratory and field settings. This should include testing multiple samples of effluent per filter per day to determine reliability of FUIs. Samples should be collected for each effluent to avoid missing data such as breakthrough of microbeads. Having longer time periods between testing dates would be aided by scheduling a longer time at the study location. Use of Quanti-trays with and without incubation would allow for more information to examine when comparing to non-incubated FUI results. Future studies may benefit from reading the Colilert test at more narrow time intervals between 36 and 48 hours to provide more information on the best balance of error probability and to identify a more optimal predictor of risk. Finally, the Colilert P/A test should be further studied with the BSF along with other indicator tests such as the water canary (Water Canary 2015) and the filter clogging assay (Hammond 2015). Follow-up on the BSFs from the field trial is recommended.

CONCLUSION

In conclusion, BSFs are a viable solution to reduce burden of disease by increasing drinking water quality, and Colilert P/A is recommended for further study as a field use indicator for biosand water filters. The Colilert P/A test deserves further investigation as a FUI. Future testing of Colilert P/A and other FUIs can also be improved, most significantly through larger samples sizes, adequate length of testing time, and inclusion of other promising tests. Much can be done to improve future evaluation of BSFs, with focus on length of follow-up and controlling for variables external to drinking water, including season.

REFERENCES

REFERENCES
Aiken
B. A.
,
Stauber
C. E.
,
Ortiz
G. M.
&
Sobsey
M. D.
2011
An assessment of continued use and health impact of the concrete biosand filter in Bonao, Dominican Republic
.
The American Journal of Tropical Medicine and Hygiene
85
(
2
),
309
317
.
Balsimo
W. V.
&
Mary
S. J. S
, .
1994
U.S. Patent No. 5,366,632. US Patent and Trademark Office, Washington, DC. Retrieved from: https://www.google.com/patents/US5366632.
Centre for Affordable Water and Sanitation Technology (CAWST)
2009
Biosand Filter Manual: Design, Construction, Installation, Operation and Maintenance
. .
Demographic and Health Survey (DHS) Program
2015
Rwanda Child Health Data 2010
.
Retrieved from: http://www.dhsprogram.com/What-We-Do/Survey-Types/DHS.cfm (accessed May 11, 2015)
.
Earwaker
P.
2006
Evaluation of Household Biosand Filters in Ethiopia
.
Silsoe Institute of Water and Environment, Cranfield University
,
UK
.
Elliot
M. A.
,
Stauber
C. E.
,
Koksal
F.
,
DiGiano
F. A.
&
Sobsey
M. D.
2008
Reductions of E. coli, echovirus type 12 and bacteriophages in an intermittently operated household-scale slow sand filter
.
Water Research
22
(
10–11
),
2662
2670
.
Fiore
M. M.
,
Minnings
K.
&
Fiore
L. D.
2010
Assessment of biosand filter performance in rural communities in southern coastal Nicaragua: an evaluation of 199 households
.
Rural and Remote Health
10
(
1483
),
1
8
.
Gill
S.
,
Hayes
J.
&
Coates
S.
2012
Diarrhoea Dialogues: From Policies to Progress
.
A call for urgent action to prevent the biggest killer of children in sub-Saharan Africa. http://www.path.org/publications/files/VAC_diarrhoea_dialogues_rpt.pdf.
Hammond
C. B.
2015
Low-cost Techniques for Assessing the Quality of Drinking Water From Biosand Filters
.
Doctoral thesis
,
Washington and Lee University
,
Lexington, VA
.
Hydraid
2014
Home page. Retrieved from: www.hydraid.org
.
IDEXX
2015a
.
IDEXX
2015b
Colilert presence/absence test procedure. Retrieved from: http://www.idexx.co.uk/pdf/en_gb/water/colilert-quanti-tray-how-to-use-sheet.pdf
.
Mangoua-Allali
A. L. C.
,
Coulibaly
L.
,
Ouattara
J. M. P.
&
Gourene
G.
2012
Implementation of biosand filters in rural area for drinking water production
.
African Journal of Food Science
6
(
24
),
574
582
.
O'Connell
B. J.
,
Slawson
D.
,
Quinn
M.
,
Scheuerman
P.
&
Ogunleye
O. O.
2017a
Review of biosand water filters
.
Waterlines
36
(
3
),
233
242
.
O'Connell
B.
,
Quinn
M.
&
Scheuerman
P.
2017b
Risk factors of diarrheal disease in children among the East African countries of Burundi, Rwanda, and Tanzania
.
Global Journal of Medicine and Public Health
6
(
1
).
O'Reilly
C. E.
,
Jaron
P.
,
Ochieng
B.
,
Nyaguara
A.
,
Tate
J. E.
,
Parsons
M. B.
,
Bopp
C. A.
,
Williams
K. A.
,
Vinjé
J.
,
Blanton
E.
,
Wannemuehler
K. A.
,
Vulule
J.
,
Laserson
K. F.
,
Breiman
R. F.
,
Feikin
D. R.
,
Widdowson
M. A.
&
Mintz
E.
2012
Risk factors for death among children less than 5 years old hospitalized with diarrhea in rural western Kenya, 2005–2007: a cohort study
.
PLoS Medicine
9
(
7
),
e1001256
.
Prüss-Ustün
A.
,
Bartram
J.
,
Clasen
T.
,
Colford
J. M.
,
Cumming
O.
,
Curtis
V.
,
Bonjour
S.
,
Dangour
A. D.
,
De France
J.
,
Fewtrell
L.
,
Freeman
M. C.
,
Gordon
B.
,
Hunter
P. R.
,
Johnston
R. B.
,
Mathers
C.
,
Mäusezahl
D.
,
Medlicott
K.
,
Neira
M.
,
Stocks
M.
,
Wolf
J.
&
Cairncross
S.
2014
Burden of disease from inadequate water, sanitation and hygiene in low-and middle-income settings: a retrospective analysis of data from 145 countries
.
Tropical Medicine & International Health
19
(
8
),
894
905
.
Samaritan's Purse
2015
Water, Sanitation, and Hygiene
.
Sisson
A. J.
,
Wampler
P. J.
,
Rediske
R. R.
&
Molla
A. R.
2013
An assessment of long-term biosand filter use and sustainability in the Artibonite Valley near Deschapelles, Haiti
.
Journal of Water, Sanitation, and Hygiene for Development
3
(
1
),
51
60
.
Stauber
C. E.
,
Printy
E. R.
,
McCarty
F. A.
,
Liang
K. R.
&
Sobsey
M. D.
2011
Cluster randomized controlled trial of the plastic biosand water filter in Cambodia
.
Environmental Science & Technology
46
(
2
),
722
728
.
Water Canary
2015
Real-time Water Quality Information for A Healthier World
.
World Health Organization
2011
Guidelines for Drinking-Water Quality
.
World Health Organization
2013
Diarrhoeal Disease
.